Benchmark
BLINK - Visual_Similarity
Vision benchmarks measure how well models understand images. Scores are percentages.
All vision benchmarksScores from EvalAI data.
- Models measured
- 4
- Scores recorded
- 4
- Sources
- 1
- IQ weight
- No weight published
Score distribution
- 56.1%
- Mean
- 55.9%
- Median
- 9%
- Std dev
- 46.3%
- 10th percentile
- 70.6%
- 90th percentile
- 46.3%
- Min
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Claude 3 OpusVisual_Similarity· EvalAI data | 70.6% |
| 2 | Gemini ProVisual_Similarity· EvalAI data | 55.9% |
| 3 | Qwen-VL MaxVisual_Similarity· EvalAI data | 51.5% |
| 4 | LLaVA 1.6 34BVisual_Similarity· EvalAI data | 46.3% |